6. Cluster Alias Failover
Introduction
This chapter describes how to implement cluster alias failover for your VMScluster.
Cluster alias failover allows you to set up an alias node to provide continued connection in case a system fails. The alias accepts incoming connection requests in a VMScluster for a server if the servicing node goes down. Use cluster alias failover primarily for NFS over UDP, but you can also use it for other TCP/IP protocols such as FTP or TELNET.
For cluster alias failover to work properly, make sure that you have the same set of directories exported from each VMScluster node. If it works effectively, cluster alias failover allows users to continue working productively if cluster nodes go down.
How It Works
Each VMScluster node has a unique internet address. With cluster alias failover, you can assign a common secondary internet address, or alias. This alias is always an Internet address and never a name. This same secondary address is given to each VMScluster node that handles connection requests, and is recognized as a local address (see Figure 6-1).
One of the nodes accepts the OpenVMS cluster-wide resource lock, adds the alias, and handles the incoming connection requests. Other VMScluster nodes are also assigned the alias queue for the resource lock. If the servicing node goes down (or you shut down TCPware on it), the system releases the resource lock. One of the queued VMScluster nodes acquires the resource lock and adds the secondary address.
CAUTION! Do not use cluster alias addresses as Domain Name Services addresses.
Setting It Up
Use the ADD SECONDARY command in the NETCU Command Reference to set up cluster failover. For example:
NETCU>ADD SECONDARY 192.168.3.101
The address 192.168.3.101 becomes the local alias address for the interface.
You can include the /CLUSTER_LOCK qualifier with the ADD SECONDARY command. This qualifier instructs the VMScluster node to accept the OpenVMS cluster-wide resource lock before adding the secondary address. If another node in the VMScluster holds the lock, the node queues for the lock and adds the address when it acquires it.
Use the REMOVE SECONDARY command in the NETCU Command Reference to remove an alias added through ADD SECONDARY. If the system holds a cluster lock, use the /ABORT qualifier to force removal of the secondary address.
Be sure to add the ADD SECONDARY command to the TCPware ROUTING.COM file so that it can take effect each time you start TCPware for OpenVMS.
Limitations
There is no concept of a primary node with cluster alias failover. The alias address only moves to another address when the active servicing node goes down. The alias does not go back to the original servicing node when it comes back up.
You can move the alias address to a particular node by issuing the REMOVE SECONDARY /ABORT command in NETCU, as follows:
1 You can also add any of the NETCU START/IP qualifiers supported for SLIP lines on the NETCU START/IP line.
The following is a sample outgoing SLIP line startup command:
NETCU START/IP SLIP-0 192.168.95.6 TXA7
2 This time do a REMOVE SECONDARY /ABORT on the node currently holding the lock. The desired node now holds the lock.
3 Do an ADD SECONDARY on all other nodes.
Using cluster alias failover has several other limitations:
• Cluster alias failover does not interact with dynamic TCP/IP load balancing (Chapter 1).
• TCP connections, such as TELNET and FTP, go down if the servicing node crashes.